WO2023132155A1

WO2023132155A1 - Prediction line calculating device, and prediction line calculation method

Info

Publication number: WO2023132155A1
Application number: PCT/JP2022/043869
Authority: WO
Inventors: 信彦佐々木; 千枝藪谷; 昌彦飯島
Original assignee: 株式会社日立ハイテク
Priority date: 2022-01-05
Filing date: 2022-11-29
Publication date: 2023-07-13

Abstract

The objective of the present invention is to provide an automated analyzing device capable of obtaining information to be used for detecting an analysis abnormality in the course of an analysis step, without analyzing in advance a specimen having a known concentration. An automated analyzing device according to the present invention acquires a standard reaction process model from among a plurality of reaction processes, and generates a prediction line by applying a data point in the standard reaction process model to measurement result data of the reaction process (see Fig. 5).

Description

Prediction line calculation device, prediction line calculation method

The present invention relates to a technique for calculating prediction lines of results measured by an automatic analyzer that analyzes the components of samples.

Automated analyzers for clinical testing use a sample dispensing mechanism to dispense samples such as blood and urine from sample containers into reaction containers, and then dispense reagents from reagent containers into the reaction containers into which the samples have been dispensed. The reagent is dispensed and stirred by the dosing mechanism. After that, the concentration of the target item contained in the sample is calculated by reacting for a certain period of time and measuring the absorbance, luminescence, etc. obtained from the reaction solution.

While the sample and the reagent are reacting, absorbance and luminescence are memorized in chronological order (e.g. absorbance and luminescence are acquired 38 times in 10 minutes and memorized in chronological order (hereafter referred to as reaction described as a process)), and the presence or absence of unintended changes in the time-series changes is determined. If there is an unintended change, the automatic analyzer can inform the user that there was an abnormality in the analysis of the item.

Patent Document 1 discloses an automatic analyzer that determines the presence or absence of an abnormality from the deviation between the absorbance or luminescence amount obtained for the Nth time and the approximated curve applied in advance to the reaction process.

Patent Document 2 acquires and stores a chemical reaction model from a sample of known concentration such as a quality control sample, calculates a reference reaction process from the amount of change in absorbance or the rate of change in absorbance of the sample, and calculates the reference reaction process and the sample. Disclosed is an automatic analyzer that determines the presence or absence of an abnormality from the divergence between the reaction process of

JP 2010-271095 A JP 2004-347385 A

In Patent Documents 1 and 2, the presence or absence of an abnormality cannot be determined until the analysis is finished. In order to quickly determine the presence or absence of an abnormality, it is desirable to obtain information that can be used to determine the presence or absence of the abnormality in the analysis process.

Also, in Patent Document 2, it is necessary to prepare a chemical reaction model in advance using a sample of known concentration. However, the time required to prepare a sample of known concentration, the time required to analyze the sample of known concentration, the cost of reagents, and the like are burdensome in clinical testing. Therefore, it is desirable to obtain information that can be used to determine the presence or absence of abnormalities without using a sample of known concentration.

The present invention has been made in view of the above problems, and is capable of obtaining information used for detecting analysis abnormalities during the analysis process without previously analyzing a sample of known concentration. An object is to provide an analyzer.

The automatic analyzer according to the present invention acquires a standard reaction process model from a plurality of reaction processes, and generates a prediction line by adding data points in the standard reaction process model to measurement result data of the reaction process. .

According to the automatic analyzer of the present invention, it is possible to obtain information used for detecting analysis abnormalities during the analysis process without previously analyzing a sample of known concentration. Problems, configurations, effects, etc. other than those described above will be clarified by the following description of the embodiments.

1 is an overall configuration diagram of an automatic analyzer 100 according to Embodiment 1. FIG. FIG. 2 is a schematic diagram showing a connection relationship between an automatic analyzer 100 and a server 200; It is an example of a standard reaction process model. 4 is a flowchart for explaining a procedure for creating a standard reaction process model by a computing unit 210; It is a figure explaining the procedure of creating the prediction line of a reaction process. It is a figure which illustrates the result of having created the prediction line using abnormal measurement result data. 7 is a graph showing a result of superimposing a prediction line created using normal measurement result data together with measurement result data. 7 is a graph showing a result of superimposing a prediction line created using abnormal measurement result data together with measurement result data. 4 is a flowchart for explaining a procedure in which a calculation unit 210 determines the accuracy of measurement result data using prediction lines; 4 is a flowchart for explaining a procedure in which a calculation unit 210 determines the accuracy of measurement result data using prediction lines; It is the example which plotted the time-dependent change of the degree of divergence in S909. A standard reaction course model for two measurements is shown. 4 is a graph illustrating changes in water content over time when subject A used a sample of skin care product A. FIG. 10 is a graph illustrating changes in water content over time when subject B used a sample of skin care product A. FIG.

<Embodiment 1>
FIG. 1 is an overall configuration diagram of an automatic analyzer 100 according to Embodiment 1 of the present disclosure. The automatic analyzer 100 mainly includes a sample disk 101, a reaction disk 104, a sample dispensing mechanism 106, a reagent disk 107, a reagent dispensing mechanism 110, a sound wave irradiation mechanism 111, a stirring mechanism 112, a photometry mechanism 114, and a reaction vessel cleaning mechanism 115. , and an overall control unit 121 . A sample container 103 containing a sample 102 is concentrically arranged on the sample disk 101 . Reaction vessels 202 are concentrically arranged on the reaction disk 104 , and the arranged reaction vessels 202 are kept at a constant temperature by the constant temperature bath circulating liquid 113 . Reagent containers 109 containing various reagents 108 are arranged concentrically on the reagent disk 107 .

The overall control unit 121 includes a control circuit 116, a photometry circuit 117, and a computer 118. The computer 118 includes an input unit 119 (eg, pointing device, keyboard, etc.), a graphical user interface (GUI) for measurement results and various operations. ) and the like are connected. In the example of FIG. 1, the control circuit 116 of the overall control unit 121 is connected to each component to control the entire device. You can also

The analysis by the automatic analyzer 100 is mainly performed as follows. First, the sample 102 placed on the sample disk 101 is dispensed from the sample vessel 103 to the reaction vessel 202 on the reaction disk 104 by the sample dispensing mechanism 106 . After that, the reaction container 202 containing the sample 102 is moved to the reagent dispensing position by the rotation of the reaction disk 104, and the reagent dispensing mechanism 110 transfers the reagent 108 used for analysis from the reagent container 109 to the sample 102. Dispense into reaction vessel 202 containing. A mixed liquid of the sample 102 and the reagent 108 contained in the reaction container 202 is called a reaction liquid 122 . Subsequently, after the reaction liquid 122 in the reaction vessel 202 is degassed by the sound wave irradiation mechanism 111 , the reaction liquid 122 in the reaction vessel 202 is stirred by the stirring mechanism 112 . The reaction vessel 202 is maintained at a constant temperature, for example, 37° C., by the constant temperature bath circulating liquid 113 filled in the lower part of the reaction disk 104, thereby promoting the reaction and stabilizing the progress of the reaction.

When the reaction liquid 122 in the reaction container 202 passes through the photometry mechanism 114 as the reaction disk 104 rotates, its optical characteristic change is measured via the photometry circuit 117 . The photometric data thus obtained are sent to the computer 118, and the concentration of the target component in the sample is obtained by the calculation unit 123 in the computer 118, and the obtained data are stored in the data storage unit 124. , the results are displayed on the output unit 120 . After the reaction, the reaction vessel 202 is cleaned by the reaction vessel cleaning mechanism 115 and either used repeatedly for the next reaction or disposed of in a reaction vessel disposal section (not shown).

FIG. 2 is a schematic diagram showing the connection relationship between the automatic analyzer 100 and the server 200. As shown in FIG. Since there may be a plurality of automatic analyzers 100, they are distinguished by alphabetic suffixes in FIG. The automatic analyzer 100A transmits measurement result data (data describing the results of measuring the reaction process of the sample) to the server 200 via the overall control unit 121A, and the automatic analyzer 100B transmits the data via the overall control unit 121B. It transmits measurement result data to the server 200 .

The server 200 (prediction line calculator) accumulates measurement result data collected from the automatic analyzer 100 . Measurement result data includes all absorbance and luminescence amounts obtained in one analysis (e.g., 38 times), measurement items, reagent manufacturing lot numbers, concentration values of measurement results, and existing automatic analyzer 100 recognizable. It suffices if it has the presence or absence of an abnormality (data alarm), and incidental information such as patient information is unnecessary.

The calculation unit 210 included in the server 200 determines the accuracy of the measurement result data by performing the procedure described later using the accumulated measurement result data. The arithmetic unit 210 can be configured by hardware such as a circuit device that implements the function, or by executing software that implements the function by an arithmetic unit such as a CPU (Central Processing Unit). can.

Figure 3 is an example of a standard reaction process model. A standard reaction process model is data describing measurement results obtained in a standard reaction process of a sample. The calculation unit 210 generates a standard reaction process model for each sample component concentration according to the procedure described later. FIG. 3 illustrates four standard reaction process models 301-304.

FIG. 4 is a flow chart explaining the procedure for the calculation unit 210 to create a standard reaction process model. This flowchart can be implemented, for example, at a monthly model update event. Each step in FIG. 4 will be described below.

(Fig. 4: Steps S401 to S403)
The calculation unit 210 acquires measurement result data accumulated in the server 200 (S401). The calculation unit 210 extracts the reaction process (time-series measurement results) for each measurement item described in the measurement result data (S402). The calculation unit 210 arranges the measurement result data, for example, for each concentration of the sample component (S403). For example, approximately 100 pieces of measurement result data are associated with one sample concentration.

(Fig. 4: Step S404)
The calculation unit 210 calculates an average value between measurement result data of similar sample concentrations. For example, assume that 100 reaction process data classified into concentrations A are collected. Each reaction process, as illustrated in FIG. 3, is described, for example, starting from the first absorbance measurement result to the 38th absorbance measurement result. The calculation unit 210 averages the values of the 100 reaction processes for the first measurement result. Similarly, 100 reaction processes are averaged for the 2nd to 38th measurement results. This results in an averaged reaction course for each sample component concentration.

(Fig. 4: Step S404: Supplement)
If there is a deviation of a threshold value or more between the reaction process obtained by the averaging process and the measurement result data, the calculation unit 210 may exclude the measurement result data from the targets of the averaging process as an outlier. The threshold is, for example, a standard deviation that can be calculated when averaging. The calculation unit 210 may perform another averaging process on the measurement result data group from which the outliers are excluded. This provides a good mean value (ie a better standard reaction process model).

(Fig. 4: Step S405)
The calculation unit 210 saves the reaction process for each concentration obtained as a result of the averaging process in S404 as a standard reaction process model. For example, the standard reaction process model for concentration A is standard reaction process model 301 in FIG. Similarly, standard reaction process models are stored for concentration B (standard reaction process model 302), concentration C (standard reaction process model 303), concentration D (standard reaction process model 304), and so on.

Fig. 5 is a diagram explaining the procedure for creating a prediction line for the reaction process. If the measurement result data is normal, the change over time of the measurement result after a certain measurement point is predicted to be similar to that of the standard reaction process model. The calculation unit 210 creates a reaction process prediction line by connecting the data points of the measurement results within a certain measurement point range to the data points of the standard reaction process model at subsequent measurement points. In the example shown in FIG. 5, the procedure for creating prediction lines for the 20th and subsequent measurement points is shown.

The computing unit 210 identifies the standard reaction process model that best matches the 20th to 23rd data points 501 . This standard reaction process model is tentatively referred to as standard reaction process model 1 . The calculation unit 210 creates a prediction line 511 by connecting the 24th to 38th data points of the standard reaction process model 1 to the data points 501 .

The computing unit 210 identifies the standard reaction process model that best matches the 20th to 24th data points 502 . This standard reaction process model is tentatively referred to as standard reaction process model 2 . The calculation unit 210 creates a prediction line 512 by connecting the 25th to 38th data points of the standard reaction process model 2 to the data points 502 .

The computing unit 210 identifies the standard reaction process model that best matches the 20th to 25th data points 503 . This standard reaction process model is tentatively referred to as standard reaction process model 3 . The calculation unit 210 creates a prediction line 513 by connecting the 26th to 38th data points of the standard reaction process model 3 to the data points 503 .

The computing unit 210 identifies the standard reaction process model that best matches the 20th to 26th data points 504 . This standard reaction process model is tentatively referred to as standard reaction process model 4 . The calculation unit 210 creates a prediction line 514 by connecting the 27th to 38th data points of the standard reaction process model 4 to the data points 504 .

When a prediction line is created for normal measurement result data by the above procedure, standard reaction process models 1 to 4 are all the same, and prediction lines 511 to 514 are also almost the same reaction process.

FIG. 6 is a diagram illustrating the result of creating a prediction line using abnormal measurement result data. In this example, computation unit 210 creates the following prediction lines, similar to FIG. 5: (a) creates prediction line 611 for data point 601; (c) generate prediction line 613 for data point 603; (d) generate prediction line 614 for data point 604;

When a prediction line is created for abnormal measurement result data by the above procedure, at least one of the standard reaction process models 1 to 4 is different from the other standard reaction process models, and the prediction lines 611 to At least one of 614 has a different reaction process from the other prediction lines.

FIG. 7 is a graph showing the result of superimposing the prediction line created using the normal measurement result data together with the measurement result data. Measurement result data 701 (solid line) indicates the actual measurement result. A prediction line 702 (dotted line) is each prediction line created by the procedure described in FIGS. The calculation unit 210 presents the result of overwriting these on one graph. The presentation format may be, for example, screen display on a display provided in the server 200, or output data describing overwritten graphs to an appropriate medium. It can be seen that the prediction lines created using normal measurement result data are grouped within a narrow range as shown in FIG.

FIG. 8 is a graph showing the result of superimposing the prediction line created using the abnormal measurement result data together with the measurement result data. It can be seen that the prediction lines created using the abnormal measurement result data are scattered over a wide range as shown in FIG.

As shown in FIGS. 7-8, the prediction lines are clustered within a narrow range (i.e., each prediction line roughly represents the same reaction process) or scattered over a wide range (i.e., each prediction line It is possible to evaluate whether the measurement result data is normal or not, depending on whether the reaction processes are different from each other). Furthermore, by overwriting as shown in FIGS. 7-8, the evaluation can be performed visually.

FIGS. 9A and 9B are flowcharts for explaining the procedure by which the calculation unit 210 determines the accuracy of the measurement result data using the prediction line. This flowchart can be implemented in order to determine the accuracy of the measurement result data after the automatic analyzer 100 acquires the measurement result data of the actual sample. Each step in FIG. 9A will be described below.

(FIG. 9A: Step S901)
The calculation unit 210 acquires sample measurement result data from the automatic analyzer 100 . The calculation unit 210 extracts standard reaction process data for each measurement item. A measurement item corresponds to, for example, a component contained in a sample.

(FIG. 9A: Step S902)
The calculation unit 210 sets one of the data points in the measurement result data as (a) the prediction start point, (b) the prediction end point, and (c) the base point. The number of data points to be used as these points can be determined in advance.

(Fig. 9A: Step S902: Supplement)
The prediction start point is the start point of the data points of the measurement result data compared when creating the prediction line, and is the 20th point in the examples of FIGS. The prediction end point is the end point of the data points of the measurement result data compared when creating the prediction line, for example, the 30th point. If it is known in advance that points 1-19 have a low correlation between concentrations and measurements, these data points are unnecessary in constructing the prediction line. Similarly, if the 31st to 38th points are also low in correlation, these data points are unnecessary. To exclude these data points, we decided to set a prediction start point and a prediction end point. The base point is, for example, the 23rd point. Base points are used in steps described later.

(FIG. 9A: Step S903)
The calculator 210 compares the data points in the measurement results to a standard reaction course model. The comparison procedure is the same as that described with reference to FIGS. However, a simplified comparison procedure can also be used, as described below.

(Fig. 9A: Step S903: Supplement 1)
In the example described in FIGS. 5 and 6, it was explained that the number of data points to be compared with the standard reaction process model is gradually increased from the 20th point as the starting point, but the number of data points to be compared with the standard reaction process model is limited to this. not a thing For example, only the most recent points may be compared with a standard reaction course model. For example, the calculation unit 210 may compare the latest measured value of the N-th point and the measured value of the (N-1)-th point one before that with the standard reaction process model. FIG. 9A shows such an example.

(Fig. 9A: Step S903: Supplement 2)
However, if the number of data points to be compared with the standard reaction process model is suppressed, the accuracy of matching with the standard reaction process model may decrease, and multiple standard reaction process models may be extracted as matching candidates. In such a case, the standard reaction process model with the highest matching accuracy should be selected through the following S904 to S906.

(FIG. 9A: Step S904)
The calculation unit 210 calculates the number of the standard reaction process model closest to the specimen (model N _B ). The reason why the standard reaction process model is specified by the reaction process up to the base point is that it is known in advance that the measurement results up to this point have the highest correlation with the standard reaction process model. _NB is used in correction processing, which will be described later.

(FIG. 9B: Step S905)
The calculation unit 210 identifies the standard reaction process model _NS that best matches the measured values from the first point to the prediction start point (20th point). This corresponds to specifying the standard reaction process model 1 described in FIGS. The calculation unit 210 similarly selects the standard reaction process model that best matches the measured values of the 1st to 21st points, the standard reaction process model that best matches the measured values of the 1st to 22nd points, . Identify the standard reaction course model _NE that best matches the measurements up to point (thirtieth point).

(FIG. 9B: Step S906)
The calculation unit 210 can specify one or more of each of the models N _S to N _E in S905. If multiple models are identified, the one that best matches the measurement result data is identified. Specifically, those that best match model N _B are identified as models N _S to N _E , respectively.

(FIG. 9B: Step S907)
The calculation unit 210 creates a prediction line for each measurement point from the prediction start point to the prediction end point. Specifically, the prediction line is obtained by performing the procedure described with reference to FIG. 5 (or S903: the simplified procedure described in the supplement) for the prediction start point to the prediction end point. This step corresponds to creating the prediction lines described with reference to FIGS.

(FIG. 9B: Step S908)
The calculation unit 210 overwrites the prediction line for each measurement point and the reaction process of the sample at the 1st to 38th points on one screen. This step corresponds to creating the overwritten graph described with reference to FIGS.

(FIG. 9B: Step S909)
The calculation unit 210 sums (a) the amount of deviation between the prediction lines for each measurement point, and (b) the amount of deviation between the 38th data point and the prediction line in the reaction process of the specimen, and the total The value is calculated as the reaction accuracy for that specimen. If the reaction accuracy exceeds the threshold, an alarm is output to notify that there is an abnormality in the sample.

(FIG. 9B: Step S909: Supplement)
A deviation between the standard reaction process model and the prediction line can be used as the threshold in this step. For example, the threshold may be three times the standard deviation.

<Embodiment 1: Summary>
The server 200 (prediction line calculation device) according to the first embodiment collects a plurality of reaction process data and uses them (for example, by averaging) to acquire a standard reaction process model, and using this, Determine the accuracy of measurement result data. Therefore, the accuracy of measurement result data can be determined without preparing a specific sample such as a known concentration sample in advance. As a result, it is possible to reduce the burden on the manufacturer side for preparing a specific sample.

The server 200 according to the first embodiment creates a prediction line by connecting the data points of the standard reaction process model to the data points in the process of acquiring the data points of the measurement result data, and the prediction line and the measurement result An alert can be sent when the degree of divergence between data exceeds a threshold. Therefore, it is possible to determine the presence or absence of an abnormality during the course of the reaction before the analysis is completed. This allows the user to quickly prepare for retesting.

<Embodiment 2>
Embodiment 2 of the present invention describes a specific example of S909. Other configurations are the same as those of the first embodiment.

FIG. 10 is an example of plotting changes over time in the degree of divergence in S909. Plots 1001 to 1003 represent the reaction accuracies (the sum of the discrepancies (a) and (b) calculated in S909) when samples 1 to 3 are analyzed with reagent lot 1 for item A. FIG. Plots 1004-1007 represent the reaction accuracy when analytes 4-1 are analyzed with reagent lot 2 for Item A.

While the change over time in the reaction accuracy of plots 1001 to 1003 is small, the reaction accuracy of plot 1004, which changed from reagent lot 1 to reagent lot 2, has a large divergence from the reaction accuracy of plot 1003. If this divergence is small, it is suggested that the difference in reagent components is small even if the reagent lot is changed, so it can be said that there is no need to re-execute calibration or the like. On the other hand, if this divergence is large, it suggests that there is a difference in the reagent components when the reagent lot changes, so it is desirable to re-execute the calibration. Therefore, the calculation unit 210 recommends the user to perform calibration.

As shown in plot 1007, when the reaction accuracy exceeds the threshold, the calculation unit 210 suggests to the user the possibility of deterioration of the reagent and recommending its replacement.

If the number of items exceeding the reaction accuracy threshold exceeds the threshold for not only item A but also multiple items, the calculation unit 210 recommends replacing the consumables of the automatic analyzer 100, such as the light source and detector. .

The calculation unit 210 can also predict future changes in reaction accuracy over time by fitting the plots shown in FIG. This makes it possible to predict when the reaction accuracy will exceed the threshold.

<Embodiment 2: Summary>
The server 200 according to the second embodiment calculates the accuracy of the measurement result data by totaling the deviation between the prediction lines and the deviation between the prediction line and the measurement result data. Whether or not there is an abnormality in the measurement result data is determined according to the change over time. As a result, the accuracy of measurement results can be determined without the need for the time required to prepare samples of known concentrations for quality control in advance, the time required to analyze the samples of known concentrations, the cost of reagents, and the like. .

In the server 200 according to the second embodiment, when the reaction accuracy does not exceed the threshold, but the plot changes over time (for example, the plot 1004 in FIG. 10, the difference between the plot 1004 and the previous plot is smaller than the threshold value on the vertical axis of FIG. 10), calibration can be prompted. Also, if there is only one measurement item whose reaction accuracy exceeds the threshold, it is possible to suggest the possibility of reagent deterioration or the like. Furthermore, if there are two or more measurement items for which the reaction accuracy exceeds the threshold, replacement of the consumable parts provided in the automatic analyzer 100 can be prompted.

Although it has been described that the server 200 according to the second embodiment determines the content of the alert according to the number of items whose reaction accuracy exceeds the threshold, this number can be set arbitrarily. Therefore, it should be added that the above numerical values regarding the number are an example. However, if the number is large, replacement of the consumables is recommended, and if the number is smaller than that, it suggests reagent deterioration, etc., which is the same regardless of the number.

<Embodiment 3>
In Embodiment 3 of the present invention, an example of analyzing the water content of human skin will be described as an example of sample components analyzed by the automatic analyzer 100 . The configuration of the automatic analyzer 100 is the same as in the first and second embodiments.

Fig. 11 shows a standard reaction process model for two measurement items. A model 901 shows changes in the water content of the basal layer inside the skin of the skin model 1 . Similarly, models 902-907 show changes in water content in the basal layer of skin models 2-7. Models 1001 to 1007 show changes in water content of stratum corneum on the skin surface of skin models 1 to 7 . Therefore, model 901 and model 1001 are associated, and similarly models 902 to 907 and models 1002 to 1007 are associated, respectively.

FIG. 12A is a graph exemplifying changes in water content over time when subject A used a sample of skin care product A. FIG. In this example, measurement results are shown at the beginning of use, 5 days after the start of use, and 10 days after the start of use. Plot 1501 is the result of measuring the water content of the basal layer inside the skin. Plot 1502 is the result of measuring the water content of the stratum corneum on the surface of the skin.

The calculation unit 210 identifies the one that is closest to the plot 1501 from among the models 901 to 907 of changes in the moisture content of the basal layer inside the skin. Here, the model 907 is assumed to be the closest. Calculator 210 creates prediction line 1107 by connecting data points from day 10 onwards of model 907 to plot 1501 . Calculator 210 similarly creates prediction line 1207 by connecting data points from day 10 onwards of model 907 and corresponding model 1007 to plot 1502 .

If the linking data indicating that the skin type of the skin model 7 is compatible with the skin care product A in terms of the effect of use is created in advance, the calculation unit 210 will Subject A can be recommended to purchase skin care product A. Further, the calculation unit 210 indicates to the subject A that it is predicted that the condition of the stratum corneum on the skin surface will improve after one month with continuous use of the skin care product A, and recommends that the subject A purchase the skin care product A. You can justify it. If the subject A can actually feel that the condition of the stratum corneum on the skin surface has improved after one month by actually continuously using the skin care product A, the reliability of the above prediction is indicated. can be done.

FIG. 12B is a graph exemplifying changes in water content over time when Subject B used a sample of skin care product A. FIG. Similar to FIG. 12A, this example shows measurement results at the start of use, on the fifth day, and on the tenth day. Plot 1503 is the result of measuring the water content of the basal layer inside the skin. Plot 1504 is the result of measuring the water content of the stratum corneum on the surface of the skin.

The computing unit 210 identifies the one that best matches the plot 1503 from among the models 901 to 907. Here, the model 902 is assumed to be the closest. Calculator 210 creates prediction line 1302 by connecting data points from day 10 onwards of model 902 to plot 1503 . Computing unit 210 similarly creates prediction line 1402 by connecting data points from day 10 onwards of model 902 and corresponding model 1002 to plot 1504 .

If the link data is created in advance that the skin type of skin model 2 is not compatible with skin care product A in terms of the effects of use, and is compatible with skin care product B in terms of the effects of use, subject B will be able to use skin care product A. When the sample has been used for 10 days, the calculation unit 210 can suggest to the subject B that the subject B use or purchase the skin care product B, which has good compatibility with the use effect, instead of the skin care product A. Further, the calculation unit 210 indicates to the subject B that the condition of the stratum corneum on the skin surface is expected to improve after one month by continuously using the skin care product B, and purchases the skin care product B. can provide a rationale for recommending that If the subject B can actually feel that the condition of the stratum corneum on the surface of the skin has improved after one month by continuously using the skin care product B, the reliability of the above prediction is indicated. can be done.

As in Embodiment 1, the prediction line and measurement result data can be overwritten on one screen. In this case, it can be expressed that the prediction error is large if the deviation between the prediction lines is large, and that the prediction error is small if the deviation between the prediction lines is small.

The calculation unit 210 uses the data used when creating the standard reaction process model or the actually measured measurement result data to calculate the statistics of the probability that the next actually measured measurement result data falls between the prediction lines. You can keep it. As a result, for example, when the user uses a sample of skin care product A for 10 days, it is possible to display the probability that the future measurement result (moisture content) will fall between the prediction lines at the same time as the prediction lines are overlaid.

<Regarding Modifications of the Present Invention>
The present invention is not limited to the embodiments described above, and includes various modifications. For example, the above-described embodiments have been described in detail in order to explain the present disclosure in an easy-to-understand manner, and are not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. Moreover, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

In the above embodiment, it was explained that the server 200 determines the accuracy of the measurement result data acquired by the automatic analyzer 100. Accuracy may be determined. In this case, the overall control unit 121 can be configured in the same manner as the calculation unit 210, and the automatic analysis device 100 functions as a prediction line calculation device.

In the above embodiment, the server 200 collects measurement result data through communication from the automatic analyzer 100, but the method of collecting measurement result data is not limited to this. Also, instead of measuring the reaction process of the specimen, the reaction process of a sample of known concentration may be collected. Furthermore, an ideal reaction process may be generated by simulation or the like and collected.

In the above embodiments, the definition of explanatory variables and the method of identifying the closest standard reaction process model are presented as an example, and other definitions and methods can also be used.

In the third embodiment, the device for measuring the moisture content of human skin does not necessarily have to be the automatic analyzer 100 described in the first embodiment. The present invention can be applied to the measurement results as long as the measurement results can be acquired from a device capable of measuring at least the moisture content of human skin and analyzed.

100: Automatic analyzer 200: Server 210: Calculation unit

Claims

A prediction line calculation device for calculating a prediction line of a measurement result measured by an automatic analyzer that analyzes the components of a sample,
The automatic analyzer measures the reaction process of the sample in chronological order,
The prediction line calculation device includes a calculation unit for calculating a prediction line of measurement result data describing the results measured by the automatic analysis device,
The computing unit acquires a standard reaction process model representing a standard reaction process generated from a plurality of the reaction processes,
The calculation unit generates a first prediction line by adding data points after the first time point in the standard reaction process model to data points of the measurement result data at the first time point in the reaction process. death,
The calculation unit calculates data points after the subsequent time point in the standard reaction process model for data points of the measurement result data at one or more subsequent time points after the first time point in the reaction process. generating one or more second prediction lines by adding
The prediction line calculation device, wherein the calculation unit outputs the first prediction line and the second prediction line.
The calculation unit determines the accuracy of the measurement result data according to the degree of divergence between at least one of the first prediction line and the second prediction line and the measurement result data. Item 1. The prediction line calculation device according to item 1.
The computing unit classifies the measurement result data in each of the plurality of reaction processes into one or more groups according to the attributes of the sample,
The calculation unit creates the standard reaction process model for each group by excluding statistical outliers for each group,
The prediction line calculation device according to claim 1, wherein the calculation unit generates the first prediction line and the second prediction line for each group using the standard reaction process model for each group. .
The computing unit acquires the standard reaction process model for each attribute of the sample,
The computing unit identifies, as a first model, the standard reaction process model that matches the reaction process from a start point before the first point in the reaction process to the first point, and
The calculation unit generates the first prediction line using the first model,
The computing unit identifies, as a second model, the standard reaction process model that matches the reaction process from the start time point to the subsequent time point,
The prediction line calculation device according to claim 1, wherein the calculation unit generates the second prediction line using the second model.
The computing unit identifies, as a reference model, the standard reaction process model that best matches the reaction process from the start point of the reaction process to a reference point in the reaction process,
When there are a plurality of candidates for the first model, the computing unit adopts the one that is closest to the reference model as the first model,
5. The prediction line calculation device according to claim 4, wherein, when there are a plurality of candidates for the second model, the calculation unit adopts the one that is closest to the reference model as the second model. .
The calculation unit displays the measurement result data on a display device, and superimposes at least one of the first prediction line and the second prediction line on the same screen as the measurement result data. The prediction line calculation device according to claim 1, characterized in that:
At a specific measurement point in the reaction process, the computing unit
difference of at least one of the first prediction line and the second prediction line with respect to the measurement result data;
and,
a difference between the first prediction line and the second prediction line;
Calculate the degree of divergence between at least one of the first prediction line or the second prediction line and the measurement result data,
2. The prediction line calculation according to claim 1, wherein, when the degree of divergence exceeds a first threshold, the calculation unit outputs an alert to the effect that the accuracy of the measurement result data has not reached a reference value. Device.
The computing unit acquires the standard reaction process model for each measurement item of the sample,
The calculation unit calculates the degree of divergence for each measurement item,
8. The prediction line calculation device according to claim 7, wherein the calculation unit sets the content of the alert according to the number of the measurement items for which the degree of divergence exceeds the first threshold.
If the amount of change over time in the degree of divergence exceeds a second threshold smaller than the first threshold, the computing unit sets the alert prompting calibration of the automatic analyzer,
When the number of the measurement items for which the degree of deviation exceeds the first threshold is greater than the first number and equal to or less than the second number, the arithmetic unit prompts the alert to replace the reagent to be reacted with the sample. and set
When the number of the measurement items for which the degree of deviation exceeds the first threshold exceeds the second number, the arithmetic unit sets the alert prompting replacement of consumable parts included in the automatic analyzer. The prediction line calculation device according to claim 8, characterized in that:
The computing unit acquires the standard reaction process model for each measurement item of the sample and for each group of the sample,
The computing unit specifies in advance an association between the standard reaction process model acquired for the first measurement item of the first sample group and the standard reaction process model acquired for the second measurement item of the first sample group. keep,
The computing unit generates the first prediction line and the second prediction line for the first measurement item by specifying the standard reaction process model that matches the measurement result data acquired for the first measurement item. death,
The computing unit identifies the standard reaction process model for the second measurement item corresponding to the standard reaction process model identified for the first measurement item according to the association,
2. The calculation unit generates the first prediction line and the second prediction line for the second measurement item using the standard reaction process model specified for the second measurement item. Prediction line calculator as described.
The first measurement item is the first moisture content of the basal layer inside the human skin,
The second measurement item is the second water content of the stratum corneum on the surface of human skin,
The group of samples is a group of subjects with a similar relationship between the number of days of skin care product use and the change in the first and second moisture content;
The computing unit identifies, for each group of subjects, a correspondence relationship between the standard reaction process model representing the change in the first water content and the standard reaction process model representing the change in the second water content. 11. The prediction line calculation device according to claim 10, wherein the association is specified by:
The computing unit specifies in advance a correspondence relationship between the recommended skin care product and the group of subjects for at least one of the first measurement item and the second measurement item,
The calculation unit acquires the measurement result data for a specific subject using a specific skin care product, and uses the measurement result data to generate the first prediction line and the second prediction line,
wherein, when the first prediction line and the second prediction line indicate favorable skin care results, the calculation unit outputs data indicating that the specific skin care product is recommended;
If the first prediction line and the second prediction line show poor skin care results, the calculation unit recommends the skin care product according to the correspondence relationship between the recommended skin care product and the subject group. 12. The predictive line calculation device according to claim 11, outputting data indicating the skin care product to be used.
The computing unit screen-displays the measurement result data on a display device, and displays the first prediction line and the second prediction line on the same screen as the measurement result data by superimposing them on the same screen. 11. The predictive line calculation device according to claim 10.
The calculation unit statistically calculates a probability that the measurement result data falls within the area formed by the first prediction line and the second prediction line,
The calculation unit displays the probability that the measurement result data acquired up to a specific point in time will fall within the area after that point along with the measurement result data, the first prediction line, and the second prediction line. 14. The prediction line calculation device according to claim 13, characterized by:
A prediction line calculation method for calculating a prediction line of a measurement result measured by an automatic analyzer that analyzes the components of a sample,
The automatic analyzer measures the reaction process of the sample in chronological order,
The prediction line calculation method has a step of calculating a prediction line of measurement result data describing the results measured by the automatic analyzer,
In the step of calculating the prediction line, obtaining a standard reaction process model representing a standard reaction process generated from a plurality of the reaction processes;
In the step of calculating the prediction line, by adding the data points after the first time point in the standard reaction process model to the data points of the measurement result data at the first time point in the reaction process, 1 generate a prediction line,
In the step of calculating the prediction line, for data points of the measurement result data at one or more subsequent time points after the first time point in the reaction course, the subsequent time points in the standard reaction course model. generating one or more second prediction lines by adding subsequent data points;
The prediction line calculation method, wherein the step of calculating the prediction line outputs the first prediction line and the second prediction line.